DATA505 Final Presentation

Nicole Rodgers & Jon Garrow

Predicting Part Profits

We used a linear regression model to predict which of CravenSpeed’s parts will be most profitable over time. Overall, we found that parts like the FlexPod are likely to be most profitable for CravenSpeed. It is a small, inexpensive plastic part that is useful for many makes, models, and years.

Importance Chart

We used a random forests technique to show which features are most important to this model.

Number of Components

We converted this to a number, ranging from 0 to 31 included in each SKU.

Percent Ordered for Stock

This variable suggests retailers are ordering this product.

Product Type

A few product types are especially profitable over the period of this dataset.

Percent Drop Shipments

This feature suggests that long-term profitability has a negative correlation with drop shipments.

Main Component Finish Process

The finishing process “Oxide” was (negatively) predictive of profitability.

Listing Type

Variants tend to be less profitable.

Year Range

We created this feature to show the number of model years a product works for.

Main Component Material

Main Component Mfg Process